// BENCHMARK

PhyBench benchmark

AI model leaderboard for the PhyBench benchmark. Compare how large language models score on PhyBench, see the full ranking, and understand what this AI benchmark measures. gemini-2.5-pro currently leads with 55.01. Physics-problem reasoning benchmark — high-school and undergraduate physics questions with numeric answers.

Leaderboard

# Model Organization Score Variant Source
#1 gemini-2.5-pro Google DeepMind 55.01 cited-ring-1t official ↗
#2 GPT-5 OpenAI 48.53 thinking-high-cited-ring-1t official ↗
#3 DeepSeek-V3.1-Terminus DeepSeek 47.91 thinking-cited-ring-1t official ↗
#4 Ring-1T Ant Group 42.65 thinking official ↗
#5 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 42.61 thinking-cited-ring-1t official ↗

Frequently asked questions about PhyBench

What is the PhyBench benchmark?

Physics-problem reasoning benchmark — high-school and undergraduate physics questions with numeric answers.

How is the PhyBench benchmark scored?

PhyBench is scored using the accuracy metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on PhyBench?

As of the latest reported scores on GenAIList, gemini-2.5-pro achieves the highest result on PhyBench with a score of 55.01.

Is a higher PhyBench score better?

Yes. On PhyBench a higher score indicates better performance, so models near the top of the leaderboard are the strongest.